AI-Powered Detection System Identifies Petroleum Contamination in Edible Oils

News
Article

Researchers from Jiangsu University and Jimei University have developed an AI-powered detection system using near-infrared spectroscopy and a convolutional neural network long short-term memory (CNN-LSTM) model to accurately identify petroleum contamination in edible oils for improving food safety and quality control.

Edible oils are a hot commodity on the global marketplace, as they are used in a wide variety of products. To ensure the authenticity and quality of edible oils, analytical techniques are needed to evaluate the edible oils before they are pushed out into the market. A recent study published in Microchemical Journal introduced a new approach to detect petroleum contamination in edible oils using artificial intelligence (AI) (1). Conducted by lead authors Hui Jiang and Quansheng Chen from Jiangsu University and Jimei University, the research presents a non-invasive and highly accurate method for identifying petroleum derivatives in olive oil, a needed advancement for food safety (1).

Dark glass bottle of extra virgin olive oil on wooden table, Cooking Essentials. Generated by AI. | Image Credit: © Nasnunt - stock.adobe.com

Dark glass bottle of extra virgin olive oil on wooden table, Cooking Essentials. Generated by AI. | Image Credit: © Nasnunt - stock.adobe.com

Edible oils, including olive oil, contain many health benefits (2). Some of these benefits include its antioxidant properties (2). These health benefits are the reason why edible oils are popular among consumers. However, edible oils are susceptible to contamination by petroleum derivatives during transportation, primarily because of inadequate tank cleaning or cross-contamination in oil tank trucks (1). If substances such as lubricating oil, engine oil, kerosene, diesel oil, and white mineral oil are consumed, they pose serious health risks. The current ongoing challenge among researchers is that traditional detection methods sometimes fall short in delivering accurate results (1). Oftentimes, these methods are also time-consuming, requiring sample preparation steps before the analysis can be conducted.

The study Jiang and Chen conducted with their team attempted to demonstrate a better method that improves on existing, traditional methods. Their method involved using near-infrared (NIR) spectroscopy and advanced AI-based classification models.

The researchers detailed their method in their study. Their method is a detection system that combines NIR spectroscopy with a hybrid deep learning model integrating convolutional neural networks (CNN) and long short-term memory (LSTM) (1). This combination enhances feature extraction and classification accuracy, outperforming traditional methods like partial least squares discriminant analysis (PLS-DA) and support vector machines (SVM) (1).

Through CNN, the spectral data was processed to extract 128 feature dimensions, which were then refined and reduced to 13 features, each containing 32 channels (1). These optimized features were subsequently input into the LSTM network, allowing the system to efficiently recognize contamination patterns while avoiding issues such as overfitting and inaccurate classifications (1).

This study demonstrates a key difference between chemometrics and AI in identifying petroleum contamination. Chemometrics methods, such as PLS-DA and SVM, struggle with parameter optimization. As a result, these methods result in misclassification and reduced detection reliability (1).

In contrast, the AI-driven approach has yielded a more stable, precise, and automated method for identifying petroleum contamination in edible oils (1). The study showed that when the researchers used CNN for feature selection and LSTM for sequential pattern recognition, the researchers were able to use their model to successfully distinguish between contaminated oil samples and pure olive oil with high accuracy.

The implications of this research extend beyond olive oil. The same methodology could be adapted to detect contamination in various edible oils, including soybean, sunflower, and palm oil (1). Furthermore, the AI-based approach may pave the way for broader applications in food authentication, ensuring that consumers receive uncontaminated and high-quality products (1).

As the global demand for food safety intensifies, the integration of AI with spectroscopic techniques could revolutionize quality control in the food industry (3). The non-invasive nature of this detection method makes it particularly attractive for large-scale implementation in food production and distribution chains (1,3). Jiang and Chen’s study not only provides a powerful tool for identifying petroleum derivatives in edible oils, but it also sets a precedent for future AI-driven innovations in food analysis.

References

  1. Zhu, J.; Deng, J.; Meng, F.; et al. Identification of Petroleum Derivatives in Olive Oil by Near Infrared Spectroscopy Combined with Convolutional Neural Network and Long Short-term Memory Interpretative Analysis. Microchem. J. 2025, 209, 112874. DOI: 10.1016/j.microc.2025.112874
  2. Wetzel, W. New Study on Edible Oil Analysis Integrates FT-NIR and Machine Learning. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/new-study-on-edible-oil-analysis-integrates-ft-nir-and-machine-learning (accessed 2025-02-28).
  3. Workman, Jr., J. Edible Oil Testing: Handheld Raman Spectroscopy Offers Quick, Reagent-Free Answers. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/edible-oil-testing-handheld-raman-spectroscopy-offers-quick-reagent-free-answers (accessed 2025-02-28).
Recent Videos
Christian Huck discusses how spectroscopic techniques are revolutionizing food analysis. | Photo Credit: © Spectroscopy.
Related Content